Patients with cancer often have readily detectable numbers of tumor-specific T cells in their peripheral blood. However, these T cells are functionally defective and do not cause tumor regression. Understanding the basis for this defective response is essential to develop improved cancer immunotherapies, and to provide immunodiagnostic assays that predict T cell immunity. We have developed a gene expression based classification of memory T cell differentiation in humans and mice. We have shown that normal memory differentiation is disrupted in dysfunctional, 'exhausted' T cells in a murine model of chronic viral infection. This suggests that defective T cells in other diseases - such as cancer - may also fail to undergo the normal memory differentiation seen in T cell responses to viral infection. Our hypothesis is that memory differentiation in tumor-specific T cells is defective, and results in a failure to acquire the functions necessary for immunologic protection. To test this hypothesis we have successfully generated gene expression profiles from rare populations of antigen- specific T cells purified by tetramer-guided sorting. This now allows us to directly examine memory differentiation in tetramer+ T cells specific for tumor antigens from patients with cancer. We propose to use a genomic approach to determine whether memory T cell differentiation in human tumor- specific cells is abnormal relative to functional virus-specific T cells. Biological validation is an essential step in the gene expression based classification of disease. Therefore, in specific aim 1, we will focus on validating our previous work. We will determine whether de novo memory formation to a novel CMV vaccine in healthy volunteers recapitulates the memory signature that we have defined. In specific aim 2, we will use the memory signature to determine whether the differentiation state of tumor- specific T cells is abnormal relative to virus-specific T cells. We will focus on myeloid leukemias and on melanoma because these diseases are associated with readily detectable tumor-specific T cell responses. At the conclusion of the proposed work, we will have achieved robust validation our previous work in a clinical setting and applied it to tumor-specific T cells in patients with cancer. This proposal will provide crucial data to support the development of gene expression based predictors of immunity. These predictors are essential to the clinical development of donor pre-immunization strategies that enhance tumor immunity in HSCT recipients. Patients with cancer do not develop effective T cell responses to their cancer but the basis for this T cell defect is not known. We propose to use a genomic approach to determine whether memory T cell differentiation in human tumor-specific cells is abnormal relative to functional virus-specific T cells. This will help identify the molecular basis for defective tumor-specific immunity, and provide a novel assay to predict immune response to cancer immunotherapy. [unreadable] [unreadable] [unreadable]